Uncovering a 260 Pc Wide, 35 Myr Old Filamentary Relic of Star Formation

MNRAS 000,1–13 (2019) Preprint 14 November 2019 Compiled using MNRAS LATEX style ﬁle v3.0

Uncovering a 260 pc wide, 35 Myr old ﬁlamentary relic of star formation

Giacomo Beccari,1? Henri M.J. Boffin,1 and Tereza Jerabkova,1,2,3,4,5,6 1European Southern Observatory, Karl-Schwarzschild-Strasse 2, 85748 Garching bei Munchen¨ 2Helmholtz Institut fur¨ Strahlen und Kernphysik, Universität Bonn, Nussallee 14–16, 53115 Bonn, Germany 3Astronomical Institute, Charles University in Prague, V Holeˇsoviˇckách 2, CZ-180 00 Praha 8, Czech Republic 4 Astronomical Institute, Czech Academy of Sciences, Friˇcova 298, 25165, Ondˇrejov, Czech Republic 5 Instituto de Astrof´ısica de Canarias, E-38205 La Laguna, Tenerife, Spain 6 GRANTECAN, Cuesta de San Jose s/n, 38712 Brena Baja, La Palma, Spain

Accepted 2019 November 13. Received 2019 November 13; in original form 2019 July 04

ABSTRACT Several recent studies have shown that the Vela OB2 region hosts a complex constellation of sub-populations with ages in the range 10 to 50 Myr. Such populations might represent the best example of the outcome of clustered star formation in Giant Molec- ular clouds (GMC). We use Gaia DR2 data over an area of 40 deg radius around the open cluster Collinder 135 to extend the study of the stellar populations of the Vela OB2 region over an area of several hundreds of parsecs on sky. Detailed clustering al- gorithms combined with the exquisite astrometric quality of the GAIA catalogue allow us to detect a new cluster named BBJ 1 that shows the same age as NGC 2547 (30 to 35 Myr), but located at a distance of 260 pc from it. Deeper investigation of the region via clustering in 5D parameter space and in the colour-magnitude diagram allows us to detect a ﬁlamentary structure of stars that bridges the two clusters. Given the extent in space of such structure (260 pc) and the young age (∼35 Myr), we exclude that such population originates by the same mechanism responsible to create tidal streams around older clusters. Even if we miss a complete picture of the 3D motion of the studied stellar structure because of the lack of accurate radial velocity measurements, we propose that such structure represent the detection of a 35 Myr old outcome of a mechanism of ﬁlamentary star formation in a GMC. Key words: open clusters and associations: general – stars: formation – stars: pre- main-sequence

1 INTRODUCTION into low mass stars. However, the vast majority of stellar cores are found at intersections of ﬁlaments and they seem Since the seminal study by Lada & Lada(2003), it is gen- to exhibit primordial mass segregation (Plunkett et al. 2018; erally understood that the vast majority of stars are formed Pavl´ık et al. 2019). As an example, Joncour et al.(2018) re- in embedded star clusters. In particular, several observa-

arXiv:1911.05709v1 [astro-ph.SR] 13 Nov 2019 cently found that nearly half of the entire stellar population tions in the infrared and sub-mm using the satellite Her- in Taurus is concentrated in 20 very dense, tiny and prolate schel and the Atacama Large Millimeter/submillimeter Ar- regions called NESTs (for Nested Elementary STructures). ray (ALMA) observatory(Andréet al. 2014; Hacar et al. Moreover the high-mass stars and embedded star clusters 2018) seem to indicate that most stars form in sub-parsec appear to be exclusively linked to places of high mass sur- high-density regions in molecular clouds (MCs). face density and multiple intersections of filaments and fibers Such high mass density sub-parsec regions in MCs are (Schneider et al. 2012; Hennemann et al. 2012; Hacar et al. made of a complex structure of intersecting molecular small- 2018). scale gas fibers and larger-scale filaments (Hacar et al. 2017, 2018). The density along individual star-forming filaments is By zooming out from individual sub-parsec embedded 5 3 high enough (10 particles/cm ) to gravitationally fragment clusters, or NESTs of star formation, we can see that gas and dust filament complexes can have lengths up to 100pc. For example, Großschedl et al.(2018) has recently used the ? E-mail: [email protected] position and astrometric information of the young stellar

© 2019 The Authors 2 Beccari, Boffin & Jerabkova objects (younger than 3Myr) in Orion A and available from stars on the celestial map in right ascension R.A. and dec- Gaia-DR2 to study the 3D shape and orientation of the giant lination Dec. As expected, we could recognize many aggre- molecular cloud (GMC) Orion A. They find that the true gates and clumps of stars likely associated with well know extent of Orion A is 90 pc, more than twice bigger with star clusters. We used the catalogue of open clusters’ mem- respect to the projected 40 pc. bers recently published by Cantat-Gaudin et al.(2018) to Moreover the sub-structure of molecular clouds seem identify the position of the know clusters in 3D space in the to be such that filaments & fibers can create large filament surveyed area. Surprisingly, a clump of stars roughly cen- complexes, usually called bones, which can be linked with tered at RA=06:21:08 Dec=-16:10:00 (l=224.221369, b=- galaxy-wide instabilities induced by bar or spiral arms (e.g. 13.866467) escaped the selection done by Cantat-Gaudin Li et al. 2016; Mattern et al. 2018). In the era of Gaia the et al.(2018). No star clusters are found in the close vicinity following question thus emerges: what imprint does this star in the recent work from Bica et al.(2019). The search for formation in seemingly large scale filaments leave in young star cluster-like objects in a 1◦ radius around the position of stellar populations just emerging from their natal molecular the cluster using Simbad (Wenger et al. 2000) gives negative clouds? result. Moreover no cluster appears at this position in the Wide field optical to near-infrared surveys revealed that WEBDA2 data base. when a star forming region is observed and studied at a scale of hundreds of parsecs it shows that young stars aggregate in multiple sub-clusters (Großschedl et al. 2018; Ksoll et al. 3 A NEW STAR CLUSTER: BBJ 1 2018; Zivkov et al. 2018). Very recently, the unprecedented astrometric precision reached through the data taken by the We used the DBSCAN data clustering algorithm (Ester Gaia satellite (Gaia Collaboration et al. 2016) and avail- et al. 1996) implemented as part of the python scikit-learn 3 able to the community trough the second data release Gaia project (Pedregosa et al. 2011) in order to identify the pres- DR2 (Gaia Collaboration et al. 2018), made it possible to ence of a star cluster at the position indicated above. DB- isolate and study the stellar population in clusters with un- SCAN requires the user to control two input parameters, equaled precision (Zari et al. 2018). namely a search radius (eps) and the minimum number of Along this line, the Vela OB2 association has recently points required to form a cluster (min samples). In short revealed to be one of the best test-bench to study the early the algorithm begins with an arbitrary starting point. Then, stages of clustered star formation (SF) at ages 10 to 35 Myr. the point’s eps-neighborhood is retrieved and if it contains a Using a combination of Gaia DR2 astrometric information sufficient number of points, a cluster forms. The DBSCAN and accurate photometry, Beccari et al.(2018) recently stud- algorithm does not require one to specify the number of clus- ied a 10 × 5 deg region around the Wolf-Rayet Star γ2 Vel, ters, and it can find arbitrarily shaped clusters in parameter including the well known clusters Gamma Vel and NGC2547. space (Gao 2014). They discovered an assembly of six clusters. Four of them In order to take full advantage of the astrometric in- formed coevally from the same molecular clouds 10 Myr formation provided by the Gaia DR2 catalogue, we per- ago, while NGC 2547 formed together with a newly discov- formed the clustering algorithm in 5D space, hence running ered cluster 30 to 35 Myr ago. Later Cantat-Gaudin et al. the search simultaneously in α, δ, $, µα and µδ. On top (2019a,b) expanded the study on a wider area (∼ 30×30 deg) of the selection of stars in $ mentioned above, we further within the Gum nebula. They hypothesise that the set of 10 restrict the clustering search to an area in Galactic coordi- Myr clusters including Vela OB2 have originated from a en- nates 222 < l < 225 and −15 < b < −13 and −8 < µα < −3 ergetic event like a supernova explosion, which could be the and 3 < µδ < 8. We run DBSCAN using eps=0.550 and cause of the IRAS Vela Shell. In this paper we extend the min samples=10. analysis of the stellar population to a wider region with the The clustering method allowed us to clearly identify the scope of investigating further any sign of formation history new cluster. The population of stars identified by the DB- of the stars in the region dynamically imprinted in the 10 to SCAN algorithm as bona-fide cluster’s members are shown 30 Myr complex stellar population. as black solid points on the left panel of Fig.1. As shown on the central panel of Fig.1 the stars do not only cluster in the coordinate space, but in proper motion around µα = −5 and µδ = 5 mas/yr. This fact confirms that we have 2 THE DATA SET identified an assembly of stars clustered in space and dynamically correlated. We show in the right panel of Fig.1 the We first retrieved from the Gaia DR2 archive1 a catalogue M ,G −G colour-magnitude diagram (CMD) of the stars of objects in a region of 40 degree radius around the center G bp r p belonging to the newly discovered population (black dots) of the cluster Collinder 135. Since we were particularly in- over-plotted to the stars belonging to the ∼30 Myr cluster terested into a detailed study of the population in the Vela NGC 2547 (Bossini et al. 2019; Cantat-Gaudin et al. 2018). OB2 region, we restricted the selection of the stars to the The comparison confirms that the stars selected in 5D space range of parallaxes 1.5 < $ < 3.5 (see Cantat-Gaudin et al. using the DBSCAN clustering algorithm in fact do belong to 2018). Moreover, following the prescription in André(2017), a ∼30 Myr single stellar population. We call the new cluster we restrict the study to only the objects for which the un- BBJ 1. The distribution of parallax of the stars belonging certainty of the parallax is lower than 10%. Initially, we visually inspected the distribution of the 2 http://www.univie.ac.at/webda/ 3 http://scikit-learn.org/stable/modules/generated/ 1 https://gea.esac.esa.int/archive/ sklearn.cluster.DBSCAN.html

MNRAS 000,1–13 (2019) 35 Myr old ﬁlamentary relic resolved by Gaia 3

Field Field 35 Myr 2 BBJ 1 BBJ 1 ZAMS 10 18 NGC 2547 BBJ 1 4 17 8 ] r y / 6 s G 16 a M m [

6 DEC [deg.] 8 15

14 10

2 94 95 96 97 12 10 8 6 4 0.5 1.0 1.5 2.0 2.5 3.0 3.5 R.A. [deg.] [mas/yr] Gbp Grp

Figure 1. Left-panel: Map of the stars in the region around the new cluster. Black dots indicate the stars identified as cluster members via the clustering algorithm while in grey we show the field stars Central-panel: The distribution of the stars in the proper motion plane. The colour code is the same as on the left panel. Right-panel: Comparison on the CMD of the stars belonging to BBJ 1 (black dots) and NGC 2547 (grey dots). The two clusters are clearly coeval. to BBJ 1 peaks at $ ∼ 2.67 ± 0.067 i.e. ∼374 pc. This fact (see Fig.2). We also restrict the clustering research to the places the cluster at the same distance of the several clusters stars in the range of proper motions −14 < µα < −3 and and association somehow related to the Vela OB2 complex, 2 < µδ < 11 mas/yr, while keeping the original parallaxes including NGC 2547 (384pc Cantat-Gaudin et al. 2019b). range 2 < $ < 3.5. We run the DBSCAN clustering algorithm using eps=0.135 and min samples=30. In the lower panel of Fig.3 we show the distribution on 4 FILAMENTARY STRUCTURES IN THE sky of the stars isolated by the initial clustering search (black VELA OB2 COMPLEX points). We indicated on the map the location of the clusters found by Cantat-Gaudin et al.(2018) in the same parallax, As described in the Introduction, the Vela OB2 complex has vα and vδ ranges. Please note that vα and vδ are expressed been the subject of several detailed investigations. In par- in Km/s and were estimated as vα = µα × 4.74/parallax and ticular Cantat-Gaudin et al.(2019b) found a complex set of vδ = µδ × 4.74/parallax. The two upper panels show the stellar populations whose spatial distribution across 180 pc distribution of the proper motions and the CMD of the same traces the IRAS Vela Shell and shows clear signs of expan- stars shown in the map. sion. The newly discovered cluster is located at a distance The first step allows us to fully recover all the clusters of ∼259 pc from NGC 2547 and ∼158 pc from Collinder already known in the region, including the new cluster BBJ 135 (see Fig.2). In the figure we show as a dash-dotted 1. The distribution of the proper motions clearly show a set line the region studied by Cantat-Gaudin et al.(2019b). As of clusters that are visible in the region. Intriguingly, while shown in Großschedl et al.(2018), these two clusters, to- the population of field stars is still strongly present in the gether with Collinder 140, NGC 2541B and UBC 7, belong catalogue as indicated by the diffuse distribution of stars to a ∼30 Myr stellar complex named Population IV which in- in the proper motion and sky space, the CMD show the cludes 5 known clusters with common distribution in proper presence of a population stars at age similar to NGC 2457 motions and distributed over an are of almost 150 pc on sky. (and the newly discovered BBJ 1) and possibly a younger The same authors note that the whole system of clusters ap- component. An old population of Red Giants is also visible pear to be bridged by a diffuse and elongated population of in the plot, certainly composed of stars belonging to the stars that physically connects the mentioned clusters. field. Moreover, a close inspection of the sky map suggests the presence of an elongated structure connecting the stel- 4.1 Clustering: Step 1 lar population in BBJ 1 to the location of Collinder 140 Admittedly, BBJ 1 is located at lower Galactic longitude and Collinder 135. On the same vein, a population of stars l with respect to the Vela complex as shown in Fig.2. seems to bridge the population of the cluster NGC 2547 to Nonetheless, driven by the intriguing observational fact that Collinder 140. the stars in BBJ 1 and NGC 2547 are coeval, we decided to explore the exquisite astrometric measurements provided by 4.2 Clustering: Step 2 the Gaia DR2 to detect the presence of a filamentary structure of stars possibly bridging BBJ 1 to the whole com- In order to investigate in depth the nature of the distribu- plex of clusters in the Vela OB2 region. We decided to use tion of the stars in the map and in the CMD, we perform again the DBSCAN clustering algorithm. We divided the a second clustering search with DBSCAN on the catalogue search in three steps. We first looked for star clusters in the obtained after the first step. The main goal of the second area included in the Galactic coordinates 210 < l < 280 and step is to remove most of the contamination from field stars −30 < b < 10, where all the targeted clusters are located that are clearly still present in the catalogue of stars ob-

MNRAS 000,1–13 (2019) 4 Beccari, Boﬃn & Jerabkova

Cantat-Gaudin et al. (2019b) 5 Trumpler 10

Haffner 13 0 Collinder 140

5 NGC 2451B

b [deg.] BBJ 1 10

Collinder 132 15 NGC 2547 Collinder 135 260 250 240 230 220 l [deg.]

Figure 2. Galactic spatial distribution of the clusters in the region. The dash-dot line indicates the region studied by Cantat-Gaudin et al.(2019b). tained after step 1. For this scope, in this second run we 4.3 Clustering: Step 3 constrain the clustering search in a 6-dimensional box, com- While the second step of clustering certainly helped in isolat- posed of the v and v , the Galactic coordinates l and b and α δ ing the clusters in the regions, the CMD of Fig.4 clearly indi- in the G − G vs M space (i.e. the CMD). This way bp r p G cates that our selected catalogue is still contaminated by an we force the clustering algorithm to search for stars that old component (> 50 Myr). Such population, well identified are distributed as coeval stellar populations on the CMD via the ZAMS model, is mostly populated by genuine MS while showing common dynamical properties. We run this field stars belonging to the Galaxy together with the oldest second session of DBSCAN using the parameters eps=0.6 clusters in the region, i.e. Alessi 21 and NGC 2527 (∼70 Myr and min samples=30. and ∼830 Myr, respectively; see Bossini et al. 2019). We thus run one last clustering filtering (step 3) in the We show in Fig.4 the distribution in vα and vδ (top- proper motions and l and b spaces with the aim to further left), on the CMD (top-right) and in the space l and b (lower separate the stellar populations and isolate the stellar com- panel) of the stars identified at the second filtering using the ponents visible in the CMD shown in Fig.4. As in step 1, DBSCAN clustering. The grey-scale shading that is shown we use eps=0.135 and min samples=30. hereafter on the figures is obtained using a density map with We show in Fig.5 and6 the result of the last clustering a Gaussian kernel. We used the function gaussian kde in- process. In the two figures we separate the stars according cluded in SciPy library (Jones et al. 01 ) to calculate the to the age of the population. We stress here that no artifi- density map. cial cut or selection is applied but the populations are shown as they come out of the clustering filtering. We used the re- normalised unit weight error (RUWE) parameter to perform The result is quite impressive. While several sub-groups a sanity check of the accuracy of the overall astrometric so- are starting to be quite well isolated in the proper motion lution of the stars in the final catalogue (i.e. after the Step space, the distribution of the stars in the CMD is mostly 3). The RUWE is expected to be around 1.0 for sources described with only 2 isochrones of 15 Myr and 35 Myr each where the single-star model provides a good fit to the astro- (dashed and solid lines, respectively) together with the Zero metric observations4. A value significantly greater than 1.0 Age Main Sequence (ZAMS) which identifies the oldest pop- (say, >1.4) could indicate that the source is non-single or ulation still present in the catalogue. The models shown in otherwise problematic for the astrometric solution. We find the figure are taken from Bressan et al.(2012) at solar metal- that 95% and 93% of the stars in our final catalogue have licity. We adopted a single extinction of A =0.124, typical V RUWE<1.4 and <1.2 respectively, independently of their of NGC 2547 (Bossini et al. 2019). We stress here that, while age. This fact proves the solidity of the astrometric solu- a dedicated study of the differential extinction might further tions of the stars that populate the final catalogue and that constraint the possible presence of intrinsic age spreads in are used hereafter for the analysis. the studied population, the adopted isochones provide an Through step 3 we perform a remarkable work in iso- excellent service to the present scope of showing our capa- bility to isolate distinct stellar populations. As visible in the sky map, a filament-like structure of stars connecting the several clusters identified in the region is also emerging. 4 Please see section 14.1.2 of the GAIA DR2 documentation

MNRAS 000,1–13 (2019) 35 Myr old ﬁlamentary relic resolved by Gaia 5

Figure 3. Upper-left: Distribution in the vα and vδ space of the stars in the region after the ﬁrst iteration with the clustering algorithm. Note that the µα and µδ are converted from mas/yr into Km/s adopting the individual parallaxes of each star. Upper-right: CMDs of the same stars shown in the top-left panel. Lower-panel: Galactic spatial distribution of the same stars shown in the top panels. The position of the known clusters is also shown. lating coeval stellar populations in the region, a young com- We emphasise here that, as a sanity check, for the latter ponent of ∼15 Myr and an older component of ∼35 Myr. group we considered only stars with a measured parame- Moreover, as shown in the sky map of Fig.6 (lower panel) ter 0.9

MNRAS 000,1–13 (2019) 6 Beccari, Boﬃn & Jerabkova

Figure 4. Upper-left: Distribution in the vα and vδ space of the stars in the region after the second iteration with the clustering algorithm. Note that the PM are converted from mas/yr into Km/s adopting the parallax. Upper-right: CMDs of the same stars shown in the top-left panel. Lower-panel: Galactic spatial distribution of the same stars shown in the top panels. The position of the known clusters is also shown. tinction is quite low and mostly affecting the region around ing is not severely affecting the magnitudes and colours of Gamma Vel where we find the youngest population that is the studied populations. likely partially embedded in the molecular clouds. Given the As already shown above, several clusters are found to extent in sky of the 35Myr old filamentary structure, the belong to such elongated stricture. Our clustering algorithm distance of most of the discovered stellar populations from identifies 2 further clusters that we label in the figure as the GammaVel region combined with the remarkable homo- BBJ 2 and BBJ 3. The study of such populations and their geneity (in term of stellar sequences) seen on the CMD of nature goes beyond the scope of this paper. Still we mention Fig.6 and7, we are convinced that the differential redden- them as signature of the presence of clumpy sub-structures along the filamentary structure.

MNRAS 000,1–13 (2019) 35 Myr old ﬁlamentary relic resolved by Gaia 7

Figure 5. Upper-left: Distribution of the 2D velocities of ∼ 15Myr old stellar component identiﬁed in the region after the second iteration with the clustering algorithm. Upper-right: CMDs of the same stars shown in the top-left panel. Lower-panel: Galactic spatial distribution of the same stars shown in the top panels.

The whole formation is confined in a range of galactic bution in vα and vδ) not to be kinematically correlated with latitudes between −17 < b < −5. Remarkably, the young (10 the old stellar component previously discussed. to 15 Myr) stellar component is confined in 2D space in a The figure further confirm the presence of a ∼260 pc region 257 < l < 270 and −15 < b < −5 and overlaps with long filaments of stars with ages between ∼30 to ∼35 Myr. the location on sky of NGC 2547. Such population corre- Part of this population corresponds to Population IV sponds to Population VII in Cantat-Gaudin et al.(2019b). in Cantat-Gaudin et al.(2019b). It is only thanks to the Finally, the clustering algorithm isolates two very low densi- extension on sky of the current study and the discovery of ties structures in the vicinity of BH 23 (see lower plot) while BBJ 1 that the structure can be seen in its entire extension. the old cluster Trumpler 10 seems (according to the distri- It is important to note that we also did the same study with a much larger region to ensure that the coeval populations

MNRAS 000,1–13 (2019) 8 Beccari, Boﬃn & Jerabkova

Figure 6. Upper-left: Distribution of the 2D velocities of ∼ 35Myr old stellar component identified in the region after the second iteration with the clustering algorithm. Upper-right: CMDs of the same stars shown in the top-left panel. Lower-panel: Galactic spatial distribution of the same stars shown in the top panels. we found here do not extend further. We can thus confirm cluster around the Wolf-Rayet Star γ2 Vel belong to a com- that the extend we report here is the maximum one. plex constellation of sub-populations confined in a region of ∼100 pc3 of volume. Later, Cantat-Gaudin et al.(2019b) extended the study on a wider area of sky in a range of distances from ∼250 to ∼550 pc. The areas studied include 5 RELIC STELLAR FILAMENT the known clusters Trumpler 10, NGC 2451B, NGC 2547, As mentioned previously, the Vela OB2 region has been re- Collinder 135, Collinder 140 as well as the extended Vela cently the subject of several works aimed at studying the OB2 complex. Using the Gaia DR2 data and detailed clus- complexity of the stellar populations in the region. In Bec- tering analysis they could identify several spatial and kine- cari et al.(2018) we demonstrated that NGC 2457 and the

MNRAS 000,1–13 (2019) 35 Myr old ﬁlamentary relic resolved by Gaia 9

35 Myr pop. NGC 2547 2 BBJ 1 BBJ 1 5 bridge stars bridge stars NGC 2547 4 0

6 5 G M b [deg.]

8 10

15 10

260 250 240 230 220 0.5 1.0 1.5 2.0 2.5 3.0 3.5 l [deg.] Gbp Grp

Figure 7. Left-panel: Map of the 35 Myr old population isolated with Step 3 selection process (grey points). The blue, black and red points show the stars belonging to NGC 2547, BBJ 1 and a selection of stars along the stellar filament bridging the entire structure.Right- panel: Comparison on the CMD of the stars belonging to BBJ 1 (black dots) and NGC 2547 (blue star) and a sample of stars along the stellar filament (red triangle). The stars are clearly coeval. matic subgroups. Such sub-groups could be separated into of the stellar structure shown in Fig.6 and8. Using the seven main groups of ages between 10 and 50 Myr. values of proper motions available with the Gaia DR2 cat- Guided by our discovery of a new cluster, BBJ 1, alogue, we calculate the average tangential velocity of the here we expand the study of Cantat-Gaudin et al.(2019a) stellar clusters identified in the ∼ 35 Myr filaments includ- and Cantat-Gaudin et al.(2019b) to a wider area, reaching ing the stars in the region next to BBJ1 along the filament l ∼ 220. We find that BBJ 1 is coeval to NGC 2547 but it (named Bridge in Fig.7). The mean proper motion of the is located at a distance of ∼260 pc from it. We identify a whole stellar population is µl = −15.5km/s, µb = −5.8km/s bridge of stars coeval to the two clusters and forming a fil- (white arrow) and has been removed. The result is shown amentary structure. Such population was observed already in Fig.9. We highlight in blue the clusters along the fila- by Cantat-Gaudin et al.(2019b) as Population IV. In our mentary structure as highlighted in the 3D plot shown in work this population is revealed to have an extended filamen- Fig.8. The figure suggests that BBJ1 and NGC2547 were tary structure spanning almost 260pc in length. not closer to each other in the past as they seem to have an almost parallel trajectory. At the same time the whole As discussed in the Introduction, evidence is growing blue structure seems to expand, a fact that confirms what that at the early stages of their formation, pre-stellar cores was already indicated in Cantat-Gaudin et al.(2019b). Cer- form in filamentary structures and clusters seem to ap- tainly, it would be important to investigate the 3D motion pear at the intersection of dense filamentary regions in gas using detailed radial velocity measurements. Unfortunately, holds (e.g. Plunkett et al. 2018). One question would then the Gaia DR2 catalogue provides radial velocities for a few be how such structures would appear on sky several tens of stars in the whole region only. We used Eq.6 from van de Myr after the formation of the stars and once the MC is dis- Ven, G. et al.(2006) to take into account the perspective sipated. We show in Fig.8 the extent and distribution of the expansion/contraction on the tangential velocities used in old stellar component in the 3D space. A blue line connects Fig.9. We used the mean radial velocities of the stars in the major filamentary structure that is confined in range of each cluster as the value representative of the system itself. distance ∼320 to ∼420 pc from us while the distance between While we report that the overall distribution of velocities the clusters are the two extremes of the structures (BBJ 1 does not changes significantly from the one shown in Fig.9, and NGC 2547) is ∼260 pc. Cantat-Gaudin et al.(2019a) detailed radial velocities measurements are urgently needed and Cantat-Gaudin et al.(2019b) used the information avail- to allow us to perform a full 3D kinematic study of the re- able with the Gaia DR2 data to study the kinematic struc- gion. ture of the seven populations that were the focus of their study. They find signs of expansion along the Galactic X While with the data-set in hand we cannot investigate and Z axes for the whole structure. Still, Population IV that in more detail its kinematic structure with respect to what is part of our 35 Myr population show a complex kinematics already done by Cantat-Gaudin et al.(2019a,b), we suggest arrangement that likely reflect the turbulent configuration that the spatial extension (∼260 pc) of the 35 Myr coeval of the primordial molecular cloud it formed from. structure discovered in this paper cannot be explained by While such a scenario requires detailed numerical simu- the mechanism that creates stellar tidal tails and streams lations to be fully understood, we here would like to extend around older clusters (e.g. Carballo-Bello et al. 2018; Mal- this hypothesis in light of the newly discovered extension han et al. 2019). The formation of such tidally formed struc-

MNRAS 000,1–13 (2019) 10 Beccari, Boﬃn & Jerabkova

Figure 8. 3D distribution of the ∼35 Myr population found in this work. tures is in fact regulated by internal cluster dynamics, ruled snapshot of star formation that happened in a complex and by two-body relaxation that make stars gradually leave the filamentary molecular clouds, like the one recently observed cluster (Binney & Tremaine 1987) and the external influence by ALMA (e.g. Hacar et al. 2018). We call such structure a of the gravitational field of the Galaxy that accelerate their relic stellar filament. dissolution process (Spitzer & Thuan 1972). Both mecha- nisms would lead to the formation of tidally stripped stars in a time-scale of a few hundred relaxation times. 6 CONCLUSIONS The young stellar populations presented in this work In this paper we use Gaia DR2 data to study the stellar show a large-scale (260 pc), 35 Myr coeval structure that is population in a region of 40 degrees of radius around the too extended and too young to be of tidal origin. We suggest young cluster Collinder 135. The stars are selected in or- that such large scale structure bear witness of a 35 Myr old der to have parallaxes 2 < $ < 3.5 and σ$ /$ < 10%. The

MNRAS 000,1–13 (2019) 35 Myr old ﬁlamentary relic resolved by Gaia 11

5 mean motion

Trumpler 10 0

Haffner 13 5 NGC 2451B Collinder 140 b [deg.] BBJ 3 Collinder 132 10 NGC 2547 Bridge Collinder 135 BBJ 2

15 BBJ 1

260 250 240 230 220 l [deg.]

Figure 9. The tangential velocities of the clusters in the region are shown as blue and red arrows. The arrows in the figure are scaled homogeneously. The mean velocity µl∗ = −15.5 km/s, µb = −5.8 km/s (white arrow) of the 35 Myr structure has been subtracted to the motion of the clusters. In blue we highlights the position and velocity vectors of the clusters along the ∼ 35Myr filament as highlighted in Fig.8. region includes the Vela OB2 complex of clusters studied data could isolate a previously unknown 17 Myr-old stel- by Cantat-Gaudin et al.(2019a) (see also Beccari et al. 2018) lar substructure in the Orion star forming region (OSFR). but extends at larger Galactic longitudes (l ∼ 220). Guided Located at a distance of 430 pc from us, this system has a by the catalogue of cluster stellar members recently compiled filament-like shape on the sky of length of ∼90 pc. Radial ve- by Cantat-Gaudin et al.(2018), we could visually identify locity follow-up observations of the stars in the system con- the presence of an over-density of stars roughly centered at firmed that the group of stars share a common kinematic. RA=06:21:08 Dec=-16:10:00 (l=224.221369, b=-13.866467) The authors call such system the Orion relic filament sug- which we could not identify with any previously known stel- gesting that, as for the relic filament shown in this work, lar cluster or association. the young age and extent of the structure suggest an in-situ We first used the clustering algorithm DBSCAN in 5D formation of the stars in a filamentary star formation event. (α, δ, $, µα and µδ) in order to isolate the new stellar cluster. The algorithm isolates 68 stars that are likely members From a theoretical point of view, Zucker et al.(2019) of the same association based on their 3D spatial distribu- recently showed the presence of large-scale filaments of tion and proper motions (see Fig.1). We compare the CMD > 100pc size in AREPO simulations that can form in Gi- of the new cluster stars with the one obtained with the stars ant Molecular Clouds by invoking simple galactic dynamics. belonging to the young cluster NGC 2547. The two CMDs Smith et al. (submitted to MNRAS) used a set of new sim- shown in Fig.1 are remarkably similar. The two clusters are ulations which self-consistently forms molecular cloud com- also located at a similar distance from us, i.e. ∼ 380pc. This plexes at high enough resolution to resolve internal substruc- fact confirms that the stars identified in 5D space via the ture all while including galactic-scale forces. These new sim- clustering algorithm belong to a single stellar population of ulations offer new support on the formation of stars in gi- ∼30-35 Myr. We call the new cluster BBJ 1. ant molecular clouds along filamentary structure of several The comparison of the CMDs shown in the right panel tenth of parsec in length. A detailed study of the kinemati- of Fig.1 demonstrate that NGC 2547 and BBJ 1 are co- cal properties of the stellar populations found in our current eval. In their recent work, Cantat-Gaudin et al.(2019a,b) work (possibly including detailed radial velocities) will pro- showed that NGC 2547 belongs to a coeval (∼30 Myr) fam- vide valuable observational constraints to allow us to infer ily of clusters which are grouped in space at the center of the initial condition of the formation of such structures. a expanding IRAS Vela Shell. The subgroup includes the Very recently Kounkel & Covey(2019) used unsupervised clusters Collinder 135, Collinder 140, NGC 2451B and NGC machine learning techniques on the Gaia DR2 catalogue to 2547. The whole population extends in Galactic longitude identify stellar clusters, associations, and co-moving groups from l ∼ 245 to l ∼ 265 and latitude b ∼ −15 to b ∼ −5 (see within 1 kpc and |b| < 30◦. In great agreement with our re- figure 1 from Cantat-Gaudin et al. 2019b). sults, they find groups of stars with ages <100Myr with a Remarkably, Jerabkova et al.(2019) using Gaia DR2 filamentary or string-like structure. By inspecting their Fig-

MNRAS 000,1–13 (2019) 12 Beccari, Boffin & Jerabkova ure 10 and the interactive figures 5 we can confirm that the Gao X.-H., 2014, Research in Astronomy and Astrophysics, 14, same structure detected in our work is also at least partially 159 detected in their fully independent work. In agreement with Großschedl J. E., et al., 2018, A&A, 619, A106 our conclusion they also suggest that these strings are pri- Hacar A., Tafalla M., Alves J., 2017, preprint, mordial and are not formed as a result of a tidal disruption (arXiv:1703.07029) Hacar A., Tafalla M., Forbrich J., Alves J., Meingast S., Grosss- of the clusters. chedl J., Teixeira P. S., 2018, A&A, 610 To answer the question asked in the introduction: what Hennemann M., et al., 2012, A&A, 543, L3 imprint does the star formation in seemingly large scale fila- Jerabkova T., Boffin H. M. J., Beccari G., Anderson R. I., 2019, ments leave in young stellar populations just emerging from MNRAS, 489, 4418 their natal molecular clouds? The young stellar population Joncour I., DuchêneG., Moraux and E., Motte F., 2018, preprint, discovered in this work and in Orion seem to suggest the (arXiv:1809.02380) presence of stellar structures that keep a memory of their Jones E., Oliphant T., Peterson P., et al., 2001–, SciPy: Open filamentary star formation. Such assembly of stars are too source scientific tools for Python, http://www.scipy.org/ young to be associated to tidal streams. While a solid conclu- Kounkel M., Covey K., 2019, AJ, 158, 122 sion on the origin of such structure requires detailed radial Ksoll V. F., et al., 2018, MNRAS, 479, 2389 velocity measurements able to reconstruct its 3D kinematics, Lada C. J., Lada E. A., 2003, ARA&A, 41, 57 Li G.-X., Urquhart J. S., Leurini S., Csengeri T., Wyrowski F., we propose that it represents the relics stellar structure of Menten K. M., Schuller F., 2016, A&A, 591, A5 star-formation happening in filamentary molecular clouds. Malhan K., Ibata R. A., Carlberg R. G., Valluri M., Freese K., 2019, arXiv e-prints, p. arXiv:1903.08141 Mattern M., Kainulainen J., Zhang M., Beuther H., 2018, A&A, 616, A78 ACKNOWLEDGEMENTS Pavl´ık V., Kroupa P., Subrˇ L., 2019, arXiv e-prints, p. This work has made use of data from the European Space arXiv:1905.09289 Agency (ESA) mission Gaia (https://www.cosmos.esa. Pedregosa F., et al., 2011, Journal of Machine Learning Research, 12, 2825 int/gaia), processed by the Gaia Data Processing and Plunkett A. L., Fernández-López M., Arce H. G., Busquet G., Analysis Consortium (DPAC, https://www.cosmos.esa. Mardones D., Dunham M. M., 2018, A&A, 615, A9 int/web/gaia/dpac/consortium). Funding for the DPAC Schneider N., et al., 2012, A&A, 540, L11 has been provided by national institutions, in particular the Spitzer Jr. L., Thuan T. X., 1972, ApJ, 175, 31 institutions participating in the Gaia Multilateral Agree- Wenger M., et al., 2000, A&AS, 143, 9 ment. This research has made use of the SIMBAD database, Zari E., Hashemi H., Brown A. G. A., Jardine K., de Zeeuw P. T., operated at CDS, Strasbourg, France. TJ acknowledges sup- 2018, A&A, 620, A172 port by the Erasmus+ programme of the European Union Zivkov V., et al., 2018, A&A, 620, A143 under grant number 2017-1-CZ01- KA203-035562. Zucker C., Smith R., Goodman A., 2019, arXiv e-prints, p. arXiv:1910.11362 van de Ven, G. van den Bosch, R. C. E. Verolme, E. K. de Zeeuw, P. T. 2006, A&A, 445, 513 REFERENCES AndréP., 2017, Comptes Rendus Geoscience, 349, 187 AndréP., Di Francesco J., Ward-Thompson D., Inutsuka S.-I., APPENDIX A: THE CLUSTERS KINEMATICS Pudritz R. E., Pineda J. E., 2014, Protostars and Planets VI, pp 27–51 In Fig.3,4,5 and6 of the manuscript we show the dis- Beccari G., Boffin H. M. J., Jerabkova T., Wright N. J., Kalari tribution of the 2D velocities in km/s in the right ascension V. M., Carraro G., De Marchi G., de Wit W.-J., 2018, MN- (α) and declination (δ) plane. This choice is driven by the RAS, 481, L11 fact that such plain was effectively used by the DBSCAN Bica E., Pavani D. B., Bonatto C. J., Lima E. F., 2019, AJ, 157, 12 clustering algorithm to isolate the stellar populations. Here Binney J., Tremaine S., 1987, Galactic dynamics we show in Fig. A1 the tangential velocities of the clus- Bossini D., et al., 2019, A&A, 623, A108 ters in the same plane to offer a view of the 2D kinematic Bressan A., Marigo P., Girardi L., Salasnich B., Dal Cero C., consistent with the plane used in the selection process. On Rubele S., Nanni A., 2012, MNRAS, 427, 127 the same vein, we show in Fig. A2 the distribution of the Cantat-Gaudin T., et al., 2018, A&A, 618, A93 proper motions of the entire stellar population in the l and Cantat-Gaudin T., Mapelli M., Balaguer-Núñez L., Jordi C., b plane as it evolves during the selection steps. These two Sacco G., Vallenari A., 2019a, A&A, 621, A115 plots consistently indicate that the difference in velocity of Cantat-Gaudin T., et al., 2019b, A&A, 626, A17 the different populations belonging to the old filamentary Carballo-Bello J. A., Mart´ınez-Delgado D., Navarrete C., Catelan structure (blue arrows on Fig. A1) is of the order of few M., MuñozR. R., Antoja T., Sollima A., 2018, MNRAS, 474, 683 km/s. On the contrary Trumpler 10 and Haffner 13 seem to Ester M., Kriegel H.-P., Sander J., Xu X., 1996. AAAI Press, pp be disconnected from the main structure, in line with their 226–231 3D spatial location shown in Fig.8. In principle, with a dif- Gaia Collaboration et al., 2016, A&A, 595, A1 ferential velocity of few km/s some of the systems might Gaia Collaboration et al., 2018, A&A, 616, A1 have been much more close to each other at the early stages of their formation (i.e.∼ 35Myr ago). Still, according to the 2D velocity vectors shown in Fig. A1 and Fig.9 it seems dif- 5 http://mkounkel.com/mw3d/ ficult to trace the cluster back to a limited and well confined

MNRAS 000,1–13 (2019) 35 Myr old ﬁlamentary relic resolved by Gaia 13

BBJ 1

Bridge 25

BBJ 2 30 Haffner 13 ] . Collinder 132 g Collinder 140 e d [

Collinder 135 NGC 2451B 40

Figure A2. The distribution of the proper motions of the stars Trumpler 10 in the region in the l and b as they appear after each clustering step as described in Sec.4. 45 BBJ 3

NGC 2547 50 100 110 120 130 [deg.]

Figure A1. The tangential velocities of the clusters in the region are shown in the right ascension (ra) and declination (dec) plane as blue and red arrows. The arrows in the figure are scaled homogeneously. The mean velocity (white arrow), µα∗ = −12.8 km/s, µδ = 10.5 km/s, of the 35 Myr structure has been subtracted to the motion of the clusters. In blue we highlights the position and velocity vectors of the clusters along the ∼ 35Myr filament as highlighted in Fig.9. spacial location. As mentioned in the paper, detailed measurements of the radial velocities of the different populations are urgently needed to assess the realistic 3D kinematic of the entire structure.

This paper has been typeset from a TEX/LATEX ﬁle prepared by the author.

MNRAS 000,1–13 (2019)